Arrhythmia detection and classification in existing Wearable Cardiac Defibrillators (WCDs) can be challenging, as therapy decisions must be made based on surface electrocardiograms (ECGs) as opposed to the intra-cardiac signals typical to Implantable Cardioverter Defibrillators (ICDs). As such, conventional WCDs typically include patient-activated “abort” features that allow the wearer to prevent the device from delivering a defibrillation shock in the event the WCD has falsely detected an arrhythmia. Detection specificity requirements for WCDs are also greater than those of Automated External Defibrillators (AEDs) because the WCD is worn continuously, whereas the AED is only applied when a patient demonstrates obvious symptoms of an arrhythmia. Accordingly, there is a continuing need for improved wearable defibrillation systems that have high sensitivity for detecting arrhythmias requiring therapy delivery, and which also reduce the occurrence of inappropriate therapy delivery.